Our product is called Sherpa
Inspiration
As AI starts deciding what people buy, we kept thinking about one thing, how do SMEs avoid getting buried by whoever has the best SEO, agent optimisation, or deepest discounts? So instead of another ranking system, we made it work more like a real market. Buyers and sellers each have their own agents, and both make decisions based on their own interests.
What it does
Merchants can create their own AI sales agent just by talking to an onboarding agent. Sherpa can then generate supported website integrations or personalised API docs for their website. That same merchant agent can also enter a marketplace where customer agents express buying intent, merchant agents compete with offers, and Visa closes the payment in-chat.
How we built it
- OpenAI for the agentic system
- Shopify integrations for merchant onboarding
- Visa's payment stack for checkout
- A custom UI to show agents negotiating and building offers
Challenges we ran into
The biggest question was how much negotiation is too much. Agents could technically negotiate forever, which is obviously useless, so for the hackathon we limited it to one counter-offer each, and bound it by user and merchant set rules, which are both dynamic in nature..
Accomplishments that we're proud of
The UI, and how easy we made it for SMEs to participate. A merchant should not need an AI team to enter agentic commerce. They should be able to explain their business and get an agent that can actually sell for them.
What we learned
Agentic commerce is less of a chatbot problem and more of a market-design problem. The interesting part is figuring out how buyer and seller agents represent competing interests without turning everything into a race to the lowest price.
What's next for Sherpa
Turn Sherpa into an actual agentic marketplace, onboard real merchants, and test how buyers and sellers use it in the real world.
Log in or sign up for Devpost to join the conversation.